Warrior_EA/research/gf_summary.py

"""Summarise WarriorGapFade real-tick tester runs: python gf_summary.py claude_gf2_EURCHF_d0 ..."""
from __future__ import annotations
import sys
import numpy as np
sys.path.insert(0, __file__.rsplit("\\", 1)[0] if "\\" in __file__ else ".")
import grid_summary as gs  # noqa: E402

COMMON = gs.COMMON

if __name__ == "__main__":
    print(f"{'run':<24}{'n':>5}{'net':>9}{'PF':>6}{'eqDD':>7}{'win':>6}{'target':>8}{'stop':>6}{'time':>6}"
          f"{'bp/trade':>10}{'2016-20':>9}{'2021-26':>9}")
    for nm in sys.argv[1:]:
        try:
            rp = gs.report(nm)
            raw = np.genfromtxt(rf"{COMMON}\gapfade_trades_{nm}.csv", delimiter=",", skip_header=1,
                                dtype=str, encoding="ansi")
        except Exception as e:  # noqa: BLE001
            print(f"{nm:<24} unreadable: {e}")
            continue
        if raw.ndim == 1:
            raw = raw[None, :]
        net = raw[:, 7].astype(float)
        ep = raw[:, 3].astype(float)
        vol = raw[:, 4].astype(float)
        why = raw[:, 8]
        yr = np.array([int(x[:4]) for x in raw[:, 2]])
        #--- price return per trade, independent of lot size, in bp
        xp = raw[:, 6].astype(float)
        side = raw[:, 12].astype(float)
        bp = side * (xp - ep) / ep * 1e4
        early, late = bp[yr <= 2020], bp[yr >= 2021]
        print(f"{nm:<24}{len(net):>5}{net.sum():>+9.0f}{rp.get('pf', float('nan')):>6.2f}{rp.get('eqdd', float('nan')):>7.2%}"
              f"{np.mean(net > 0):>6.0%}{np.sum(why == 'target'):>8}{np.sum(why == 'stop'):>6}{np.sum(why == 'expert'):>6}"
              f"{bp.mean():>+10.1f}{early.mean() if len(early) else float('nan'):>+9.1f}{late.mean() if len(late) else float('nan'):>+9.1f}")
    print("bp/trade = fill-to-fill price move (spread paid at both fills); 'target' exits are in-EA closes"
          " counted by the journal as expert, so the split is approximate.")